How tub-rip/event_based_optical_flow works
This repository implements research workflows for event-based optical-flow estimation: load event sequences, optimize motion by contrast maximization, then evaluate and visualize predicted flow. The README also describes depth and ego-motion in the journal work, but the supplied tree and excerpts substantiate an optical-flow implementation. Edges reflect visible calls or SOURCE INDEX references; unsampled variants are included as capabilities, with uncertain runtime wiring omitted.
Main entry points: base.py Solver Base · main.py Experiment Runner · warp.py Event Warping · patch_contrast_base.py Patch Contrast
Components 23
Evaluation Workflow
- MVSEC Sequences
src/data_loader/mvsec.py - Dataset Loading
src/data_loader/base.py - Experiment Runner
main.py - Flow Evaluation
src/solver/base.py - Flow Visualization
src/visualizer.py
Motion Estimation
- Solver Base
src/solver/base.py - Patch Contrast
src/solver/patch_contrast_base.py - Mixed Solver
src/solver/patch_contrast_mixed.py - Pyramid Solver
src/solver/patch_contrast_pyramid.py - Time-Aware Solver
src/solver/time_aware_patch_contrast.py - EV-FlowNet
src/solver/nnmodels/ev_flownet.py - Neural Layers
src/solver/nnmodels/basic_layers.py - SciPy Autograd
src/solver/scipy_autograd/scipy_minimize.py - Torch Wrapper
src/solver/scipy_autograd/torch_wrapper.py - Autograd Base
src/solver/scipy_autograd/base_wrapper.py
Contrast Objectives
- Contrast Cost Suite
src/costs - Hybrid Cost
src/costs/hybrid.py
Representations & Utilities
- Event Warping
src/warp.py - Event Image Builder
src/event_image_converter.py - Feature Calculator
src/feature_calculator.py - Flow & Event Utilities
src/utils - Flow Types
src/types/flow_patch.py
Other components
- Researcher
Connections 30
- Researcher → Experiment Runner: runs experiment
README.md - Experiment Runner → Dataset Loading: selects dataset
main.py - Experiment Runner → MVSEC Sequences: loads sequence
main.py - MVSEC Sequences → Flow & Event Utilities: uses utilities
src/data_loader/mvsec.py - Experiment Runner → Solver Base: optimizes events
main.py - Experiment Runner → Flow Evaluation: evaluates flow
main.py - Experiment Runner → Flow Visualization: configures display
main.py - Solver Base → Contrast Cost Suite: sets up objective
src/solver/base.py - Solver Base → Event Image Builder: builds event images
src/solver/base.py - Solver Base → Event Warping: warps events
src/solver/base.py - Solver Base → Flow Visualization: renders results
src/solver/base.py - Solver Base → SciPy Autograd: uses optimizer
src/solver/base.py - Patch Contrast → Solver Base: extends solver
src/solver/patch_contrast_base.py - Patch Contrast → Event Warping: warps events
src/solver/patch_contrast_base.py - Patch Contrast → Flow Visualization: renders results
src/solver/patch_contrast_base.py - Mixed Solver → Patch Contrast: uses patch base
src/solver/patch_contrast_mixed.py - Mixed Solver → SciPy Autograd: uses optimizer
src/solver/patch_contrast_mixed.py - Pyramid Solver → Patch Contrast: uses patch base
src/solver/patch_contrast_pyramid.py - Pyramid Solver → Contrast Cost Suite: uses objective
src/solver/patch_contrast_pyramid.py - Pyramid Solver → Event Warping: warps events
src/solver/patch_contrast_pyramid.py - Pyramid Solver → SciPy Autograd: uses optimizer
src/solver/patch_contrast_pyramid.py - Time-Aware Solver → Flow & Event Utilities: uses utilities
src/solver/time_aware_patch_contrast.py - EV-FlowNet → Neural Layers: uses layers
src/solver/nnmodels/ev_flownet.py - SciPy Autograd → Torch Wrapper: wraps SciPy calls
src/solver/scipy_autograd/scipy_minimize.py - Torch Wrapper → Autograd Base: extends wrapper
src/solver/scipy_autograd/torch_wrapper.py - Contrast Cost Suite → Hybrid Cost: offers combined loss
src/costs/__init__.py - Event Warping → Flow & Event Utilities: uses flow transforms
src/warp.py - Event Warping → Feature Calculator: uses feature calculator
src/warp.py - Flow Visualization → Event Image Builder: builds event images
src/visualizer.py - Flow Visualization → Event Warping: uses warping
src/visualizer.py